CVS log for src/sys/dev/sysmon/sysmon_envsys.c

Up to [cvs.NetBSD.org] / src / sys / dev / sysmon

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

s/appropiate/appropriate/ in comments and readme file.

sysmon(9): Fix callout/thread synchronization.

Callout may ONLY take sme_work_mtx, at IPL_SOFTCLOCK; MUST NOT touch
sme_mtx at IPL_NONE.  All state the callout needs is serialized by
sme_work_mtx now:

- calls to sme_schedule_callout
- calls to sme_schedule_halt
- struct sysmon_envsys::sme_events_timeout
- struct sysmon_envsys::sme_events_list
- struct sysmon_envsys::sme_callout_state
- struct envsys_data::flags
  => yes, this is a little silly -- used for ENVSYS_FNEED_REFRESH
  => should maybe separate the static driver-defined features from
     the state flags needed by sysmon_envsys but not important now

Sleeping under sme_work_mtx (except on other adaptive locks at
IPL_SOFTCLOCK) is forbidden.  Calling out to the driver under
sme_work_mtx is forbidden.

This should properly fix:

https://mail-index.netbsd.org/tech-kern/2015/10/14/msg019511.html
PR kern/56592

sysmon: Delete trailing whitespace.  No functional change intended.

s/aquire/acquire/ in comments, also one typo fix acqure->acquire.

Update for proplib(3) API changes.

Update for proplib(3) API changes.

Pull up following revision(s) (requested by riastradh in ticket #1733):

	sys/dev/sysmon/sysmon_envsys.c: revision 1.145

Don't queue sysmon refresh until the rndsource is attached.

Using the rndsource, as refreshing the sensors will do, is not
allowed until _after_ rnd_source_attach.

XXX pullup-7
XXX pullup-8
XXX pullup-9

Pull up following revision(s) (requested by riastradh in ticket #1555):

	sys/dev/sysmon/sysmon_envsys.c: revision 1.145

Don't queue sysmon refresh until the rndsource is attached.

Using the rndsource, as refreshing the sensors will do, is not
allowed until _after_ rnd_source_attach.

XXX pullup-7
XXX pullup-8
XXX pullup-9

Pull up following revision(s) (requested by riastradh in ticket #946):

	sys/dev/sysmon/sysmon_envsys.c: revision 1.145

Don't queue sysmon refresh until the rndsource is attached.

Using the rndsource, as refreshing the sensors will do, is not
allowed until _after_ rnd_source_attach.

XXX pullup-7
XXX pullup-8
XXX pullup-9

Don't queue sysmon refresh until the rndsource is attached.

Using the rndsource, as refreshing the sensors will do, is not
allowed until _after_ rnd_source_attach.

XXX pullup-7
XXX pullup-8
XXX pullup-9

Sync with HEAD

in sysmon_envsys_unregister(): loop over sme_sensors_list with TAILQ_FIRST
instead of TAILQ_FOREACH.

sysmon_envsys_sensor_detach() removes the sensor from sme_sensors_list.
Hence using TAILQ_FOREACH is not safe.

Suggested by thorpej and jmcneill.  Thanks!

Sync with HEAD

Avoid dereferencing NULL if we attempt to look up an known unit type.

Sync with HEAD

Defend against some table-lookup-not-found errors.

update from HEAD

Fix fallout from ticket #1511:
It's rnd_detach_source, not rnd_detach_sources.

Fix fallout from ticket #1511:
It's rnd_detach_source, not rnd_detach_sources.

Fix fallout from ticket #1511:
It's rnd_detach_source, not rnd_detach_sources.

Apply patch (requested by pgoyette in ticket #1511):
Detach the rndsrc before re-attaching it.

Apply patch (requested by pgoyette in ticket #1511):
Detach the rndsrc before re-attaching it.

Apply patch (requested by pgoyette in ticket #1511):
Detach the rndsrc before re-attaching it.

Pull up following revision(s) (requested by pgoyette in ticket #281):
	sys/dev/sysmon/sysmon_envsys.c: 1.140-1.141
	sys/dev/sysmon/sysmon_envsys_events.c: 1.120-1.121
	sys/dev/sysmon/sysmonvar.h: 1.50
Fixes a problem that some driver(e.g. acpitz(4) or coretemp(5)) which
use sysmon_envsys sleep waiting at "rndsrc" when "drvctl -d".
Don't call rnd_detach_source() in sme_remove_event() which is called
from sme_event_unregister_all(). Instead, call rnd_detach_source() in
sysmon_envsys_sensor_detach() and call sysmon_envsys_sensor_detach()
before sme_event_unregister_sensor(). Each sensor(envsys_data) has each
rnd_src, but some sme_events point to the same rnd_src in a sensor.
Calling rnd_detach_souce() twice with the same rnd_src brokes a reference
count in rnd_src. OK'd by pgoyette@.
--
Improve tracking of the state of an event's callout, and protect all
queries or modifications of the state with the sme_mtx mutex.
Detach the rndsrc before re-attaching it (with potentially new values).
Clean up some lock-ordering issues.
And a couple of KNF issues for good measure!
Should address PR kern/52533

Improve tracking of the state of an event's callout, and protect all
queries or modifications of the state with the sme_mtx mutex.

Detach the rndsrc before re-attaching it (with potentially new values).

Clean up some lock-ordering issues.

And a couple of KNF issues for good measure!

Should address PR kern/52533

XXX Pullup-8 along with the previous fixes from msaitoh@
XXX http://mail-index.netbsd.org/source-changes/2017/09/06/msg088028.html
~
~

 Fixes a problem that some driver(e.g. acpitz(4) or coretemp(5)) which
use sysmon_envsys sleep waiting at "rndsrc" when "drvctl -d".
Don't call rnd_detach_source() in sme_remove_event() which is called
from sme_event_unregister_all(). Instead, call rnd_detach_source() in
sysmon_envsys_sensor_detach() and call sysmon_envsys_sensor_detach()
before sme_event_unregister_sensor(). Each sensor(envsys_data) has each
rnd_src, but some sme_events point to the same rnd_src in a sensor.
Calling rnd_detach_souce() twice with the same rnd_src brokes a reference
count in rnd_src. OK'd by pgoyette@.

Sync with HEAD (as of 26th Dec)

sysmon's components need to be MODULE_CLASS_DRIVER so they will get
initialized before we configure/initialize any devices that interact
with them.

Thanks, marty!

Note the sensor number in the error output.  Useful for drivers adding
multiple sensors.

Sync with HEAD

Handle early initialization requirements - thanks martin@ and others

make things boot again, from martin.

Use tabs for indentation.

Release the mutex in the error path.

The sysmon_envsys module also depends on sysmon_power (for delivery of
sensor events to userland).

Remove a left-over debugging printf()

Modularize sysmon and its components

Convert sys/dev to use <sys/rndsource.h>.

Pull up following revision(s) (requested by hannken in ticket #666):
	sys/dev/sysmon/sysmon_envsys.c: revision 1.129
	sys/dev/sysmon/sysmon_envsys_events.c: revision 1.114
	sys/dev/sysmon/sysmonvar.h: revision 1.46
Add a counter of busy events and stop enqueueing more work if a device =
is busy.
Protect this counter with a new short time lock "sme_work_mtx" and
keep "sme_mtx" as long time lock.
Removes a deadlock where an active event holds "sme_mtx", the callout
"sme_events_check" blocks on "sme_mtx" and callout processing stops.

Sync with HEAD

Add a counter of busy events and stop enqueueing more work if a device is busy.
Protect this counter with a new short time lock "sme_work_mtx" and
keep "sme_mtx" as long time lock.

Removes a deadlock where an active event holds "sme_mtx", the callout
"sme_events_check" blocks on "sme_mtx" and callout processing stops.

Pull workqueue_destroy out of sme->sme_mtx

workqueue_destroy may sleep so we shouldn't run it with holding a mutex.

Requested by riastradh@.

Rebase to HEAD as of a few days ago.

Merge tls-earlyentropy branch into HEAD.

Be a little more clear and consistent about harvesting entropy from devices:

1) deprecate RND_FLAG_NO_ESTIMATE

2) define RND_FLAG_COLLECT_TIME, RND_FLAG_COLLECT_VALUE

3) define RND_FLAG_ESTIMATE_TIME, RND_FLAG_ESTIMATE_VALUE

4) define RND_FLAG_DEFAULT: RND_FLAG_COLLECT_TIME|
   RND_FLAG_COLLECT_VALUE|RND_FLAG_ESTIMATE_TIME

5) Make entropy harvesting from environmental sensors a little more generic
   and remove it from individual sensor drivers.

6) Remove individual open-coded delta-estimators for values from a few
   places in the tree (uvm, environmental drivers).

7) 0 -> RND_FLAG_DEFAULT, actually gather entropy from various drivers
   that had stubbed out code, other minor cleanups.

resync with head

sync with (a bit old) head

Remove trailing white space(s) from name of random source.
And also, changind space(s) in the middle with '-'.
PR#47232.

sync with head

Remove stray CR characters introduced in ticket #544.  From pgoyette.

Pull up following revision(s) (requested by pgoyette in ticket #544):
	sys/dev/sysmon/sysmon_envsysvar.h: revision 1.42
	sys/dev/sysmon/sysmon_envsys_events.c: revision 1.99
	sys/dev/sysmon/sysmon_envsys.c: revision 1.119
When unregistering a sensor device, make sure we unregister and delete
all the associated events.

Remove (commented-out) definitions for debugging - thanks to macallan@
for defflagging these.

Fix indentation.  No code change.

Simplify some logic - no functional change intended.

While here, clean up a couple of comments.

Make sure to re-add the sensor as an entropy source after user request
to reset (via envstat -S)

Extend previous changes so that even sensors which don't use a refresh()
callback to update the value can be polled to provide rnd(4) entropy.

If a sensor is flagged as capable of providing rnd(4) with entropy,
hook the sensor into rnd subsystem, and make sure we periodically
refresh the sensor whether or not it is being actively monitored.

When unregistering a sensor device, make sure we unregister and delete
all the associated events.

sync with head

merge to -current.

Don't export sysmon_envsys_list as a common.

According to the Oxford Dictionary of Etymology, a wether is a male sheep
or ram.

Catchup with rmind-uvmplock merge.

Instead of crashing, print a usable error message when a driver eroneously
does not properly initialize sensor state.

Previous fix broke the setting of current value for Indicator and
Battery-Charge sensors.  Fix this.

sync with head

Factor out some duplicated code to simplify maintenance.  Reduces
the source file by ~100 lines, and amd64 object file shrinks by
~650 bytes.

No functional change intended.

Sync with HEAD.

Since nothing actually records or maintains an average sensor value,
remove value_avg from the prop_dict.  We can't completely remove it
because there's one driver that uses this field for its own private
purposes, so for now we just rename the member.

sync with head

When the user updates the sensor device's refresh timer, reset the
callout immediately rather than waiting for the previous timer to
expire.

Be sure to always initialize props variable to a sensible value.
Fix a problem with envstat -S call which was installing bogus/random
limits.

ok pgoyette.

Extract searching of description tables into a single function, rather
than duplicating the code every time.  Minor reduction in code size
(about 1200 bytes on amd64), no change in functionality.

Handle removal of limit properties correctly, even for sensors that
provide their own internal limit(s).

When removing properties (as in envstat -S), clear the flags that track
whether the properties actually exist.

When setting new limit values, don't complain/ignore about duplicate
values if the associated flag bits (indicating that the limit is valid)
are not set.

These two fixes together should fix a problem reported in private Email
by njoly@ a couple months ago.  Issue replicated and solution tested
using my recent swsensor pseudo-device running inside qemu world!

sync with head.

sync with head

Sync with HEAD.

Use a typedef to define the prototype of the per-sensor callback.

Update manpage for the prototype, and fix a fubar'd Cross-reference.

When removing sensor properties with envstat -S, ensure that drivers
which could have modified hardware state are informed so that original
state can be restored.

Welcome to 5.99.27

When unregistering a device, in addition to reducing the global count
of sensors, we need to adjust the sme_fsensor values for some devices.
Failure to do this leaves a sparesly-populated number-space and can
subsequently result in overlapping sensors number assignments.  (This
only affects the compatability-mode API, but still needs to be fixed.)

Permit creation of an event-monitor for alarm limits even if the limits
are not currently set (and, in case of battery capacity sensors, if there
is no value_max).  Ensure that such an event-monitor does not trigger the
delivery of any actual events.

Provide a mechanism for sensors to set their limits at time other than
system startup (for example, when a battery is inserted).

This allows us to boot a system with a battery missing, install the
battery some time later, and automatically monitor it without requiring
any user intervention to create the event-monitor.  (The actual changes
for battery sensors to use this new capability will come later.)

Constify the args to sysmon_envsys_foreach_sensor()'s callback function.

Fix the locking protocol in sysmon_envsys_foreach_sensor(), and add an
additional argument to control whether or not the sensors are refreshed
before invoking the callback routine.

Make this work the way the preceeding comment blocks says it works, and
break out of the while{ } when the callback returns false, not when it
returns true!

Add a routine to iterate over all the sensors on the system.  This will
be needed to update acpiapm_get_powstat() which is currently slightly
broken due to recent rearrangement of acpi_bat(4)'s sensors.  (This
approach is in lieu of exporting proplists between kernel entities, and
avoids exposing some internal sysmon_envsys details.)

Export the sensor device's class.

Do the initial refresh before creating the event monitors, in case the
event monitors require some of the data loaded by the refresh.  For
example, the initial refresh might load sensor limits and max_value in
addition to the cur_value.

Retire the 'monitor' member of the envsys_data_t.  It was only used in
one place, and functioned as a logical OR of the ENVSYS_FMON* flag bits.

sync with head

For symmetry, allow warning and critical upper-limits for capacity
sensors.

Remove the flags member of the limits structure.  Its value was only
meaningful for a few limited function calls;  all persitent storage
of limit-present flags is maintained in the edata->upropset member.

One more callback of sme_refresh() routine needs to check first that
there is a routine to be called!

Fix an outdated comment, noticed by Paul Goyette.

Add a convenience function to find the maximum value currently reported
by a set (selected via a passed callback predicate) of sensors.
This provides an easy way to query the current temperature of a thermal
zone, for example, from within the kernel - assuming the caller knows
the topology.

Validate limit-based monitoring in one place, so we can apply the
constraints to both driver-requested and user-requested monitoring.

While here, clean up some of the debugging DPRINTFs.

Sync with HEAD.

sync with head.

Remove some code (bracketed by #ifdef NOTYET / #endif) that should not
have been committed in the first place.

Remove driver name from debug message.  Not all drivers have set this
early enough, and we can easily figure out the driver name from some
subsequent debug messages.

sync with head

Add capability to send sensor limit values to the driver so they can
be programmed into device registers.  This way we can let the hardware
help us out instead of having to always compare the value against each
limit.  (Driver updates for some sensors to take advantage of this
capability will be forthcoming.)

General clean-up and some restructuring of event handling.  This is a
precursor to letting sensor drivers actually exchange limits/thresholds
with user-land.

Don't allocate the private list-threading structure unless we're going
to use it.  Otherwise we end up leaking little bits of memory for each
sensor that is not monitored at time of initial registration.

sync with head.

Sync with HEAD.

Sync with HEAD.

Sync with wrstuden-revivesa-base-2.

1. provide for warning-min and warning-max alarm levels in addition to
   critical-min and critical-max; also, a battery warning-capacity is
   added in addition to a critical-capacity.

2. usr.sbin/envstat is modified to introduce a -W command line switch to
   display the warning-* values instead of the critical-* values, and
   envstat(8) and envsys.conf(5) man pages are updated appropriately.

3. Treat user-defined limits as a single continuum and generate a single
   event regardless of how many boundaries a change in sensor value
   crosses;  ditto for driver-defined limits.

Fixes my PR/39021
Fixes my PR/39022

OK'd by christos@ bouyer@ cube@

Sync w/ -current. 34 merge conflicts to follow.

Sync with HEAD.

Also fix build.

sync with head

Add SME_POLL_ONLY flag, which informs sme to only refresh the sensor when
it is polled (and not from GTREDATA).

Sync with HEAD.

Initialize and destroy the callout in sme_events_init() and
sme_events_destroy() respectively. And don't stop twice the callout
when calling sysmon_envsys_unregister().

Introduce per-device locking/synchronization and maintain only a
global mutex for the linked list of devices and the global proplib
dictionary.

This has improved locking contention a lot when multiple devices with
multiple monitoring events are running:

New:

 0.30      35      0.33 sme_global_mtx         sysmonioctl_envsys+28b
 0.10      10      0.11 00000000cd97feac       sysmon_envsys_acquire+4c
 0.08       6      0.09 00000000cd97feac       sme_update_dictionary+24f
 0.01       4      0.01 00000000cd97feac       sme_events_worker+2f
 0.01      10      0.01 00000000cd97fe2c       sysmon_envsys_acquire+4c
 0.00       1      0.00 00000000cd97fe2c       sysmon_envsys_release+3b
28.38      94      9.16 sme_global_mtx         sysmonioctl_envsys+28b
 4.54      74      1.47 00000000cd97fe2c       sysmon_envsys_acquire+4c
 0.06       3      0.02 00000000cd97fe2c       sysmon_envsys_release+3b
 0.03       1      0.01 00000000cd97fe2c       sme_events_worker+2f
 1.40      19      0.45 00000000cd97bee4       sysmon_envsys_acquire+4c

Old:

 4.25     313      4.74 sme_mtx                <all>
 3.12     185      3.49 sme_mtx                sme_events_worker+21
 1.12     128      1.25 sme_mtx                sysmonioctl_envsys+29b
34.75    1423     59.52 sme_mtx                <all>
22.08     477     37.82 sme_mtx                sysmonioctl_envsys+29b
12.67     946     21.70 sme_mtx                sme_events_worker+21

sync with head.

Fix some problems reported by <drochner> a while ago on tech-kern:

sme_events_worker: skip sensors with invalid state.
sme_acadapter_check:

	* use a boolean to check if an SME_CLASS_ACADAPTER was found.
	* Always refresh sensor data before checking value_cur.

Introduce the SME_INIT_REFRESH flag. If a device sets this it will mean
that after interrupts are enabled in the autoconf(9) process, sensor
data will be refreshed once.

sync with HEAD

Sync with HEAD.

sync with head.

Destroy the callout via sysmon_envsys_destroy() now that it's
initialized in sysmon_envsys_create().

sync with head

sync with HEAD

Sync with HEAD

- Initialize the callout with CALLOUT_MPSAFE, like it was done before
  in sysmon_envsys_events.c.
- Separate sme_event_check_low_power() into sme_battery_check(),
  sme_acadapter_check() and sme_battery_critical() (no functional changes).

In sysmon_envsys_create(), initialize the callout structure.  Stops
panics with a DIAGNOSTIC kernel.

In sysmon_envsys_find_40(), wait until an envsys is not busy before
settings its busy flag, in the pattern of sysmon_envsys_find().

Use LIST_FOREACH().

Use TAILQ_FOREACH().

Sync with HEAD.

Sync with HEAD.

Sync with head.

sync with head

MUTEX_SPIN -> MUTEX_DEFAULT.

Sync with HEAD.

Sync with HEAD

Update a comment.

After comments from Andrew Doran:

- The mutex for the callout handler must run at IPL_SOFTCLOCK.
- Just stop the callout in sysmon_envsys_unregister() and don't wait
  for the callout to finish.

Sync with HEAD.

Sync with HEAD

Extend the envsys2 API (one more time, sorry) as defined in:

http://mail-index.netbsd.org/tech-kern/2007/11/09/0001.html

sysmon_envsys_create() and sysmon_envsys_destroy() were added to
create/destroy sysmon_envsys objects (and its TAILQ/LIST for sensors/events).

sysmon_envsys_sensor_attach() and sysmon_envsys_sensor_detach() were
added to attach/detach sensors to a specified sysmon_envsys device.

The events framework is now per device and configurable via the
ENVSYS_SETDICTIONARY ioctl or /etc/envsys.conf and envstat(8).

Update all users and documentation to reflect these changes.

sync with head.

Sync with HEAD

sync with HEAD

Sync with HEAD.

- Rename ENVSYS_BATTERY_STATE units to ENVSYS_BATTERY_CAPACITY and
  introduce ENVSYS_BATTERY_CHARGE, which is the same than an Indicator and
  it's used to know if the battery is currently charging or discharging.

- Require two sensors at least for SME_CLASS_BATTERY to make the low-power
  condition work: a ENVSYS_BATTERY_CAPACITY plus ENVSYS_BATTERY_CHARGE.

- Simplify sme_event_check_lowpower() and make it check the required
  sensors in the SME_CLASS_ACADAPTER and SME_CLASS_BATTERY classes.
  If the acadapter is not ready, trust the state returned by the battery
  device.

Based on suggestion from joerg@.

do not embed acpi device names in the sensor names
fake it in the compat code for the old envsys API

sync with head.

sync with head.

Sync with head.

Sync with HEAD.

New changes to support the new configuration file format for
envstat(8) and the envsys framework:

- Modify the ENVSYS_SETDICTIONARY ioctl to support the following
  plist structure:

  <dict>
	<key>foo0</key>
	<array>
		<dict>
			<key>index</key>
			<string>sensor0</string>
			<key>description</key>
			<string>cpu temp</string>
			...
			Another property for this sensor
			...
		</dict>
		...
		Another dictionary for other sensor
		...
	</array>
	...
	Another device as above
	...
  </dict>

  Multiple devices with multiple sensors can now be specified, that means
  that to set the properties only one copyin(9) is needed now.

- Added the ENVSYS_REMOVEPROPS ioctl, that accepts a boolean object
  "envsys-remove-props" and when set to true, all properties that were
  set previously by ENVSYS_SETDICTIONARY will be removed. That means that
  you can now set multiple critical limits, descriptions and all they
  will be removed or changed to its default value (for rfact and
  description objects).

- Added the 'index' and 'allow-rfact' objects into the sensor dictionaries,
  for better interactivity with userland. To know the position of the
  sensor and to know if sensor allows to change the rfact.

- Misc cosmetic changes for consistency.

- Use a two clause license for all my code.

sync with head.

Sync with HEAD.

sme_add_sensor_dictionary: don't forget to release the dictionary when
an error ocurrs. This is the last memleak in the code, let me know if
you find more and I'll pay you a beer.

Sync with HEAD.

Change the DPRINTFs in sysmon_envsys_destroy_plist() to DPRINTFOBJs
(second level of verbosity).

sysmon_envsys_register: don't forget to clear SME_FLAG_BUSY when
registration succeeds.

Use DPRINTF for a debugging printf.

Remove an unused variable.

Fix another memleak that only happened when using
sysmon_envsys_unregister(): in sysmon_envsys_destroy_plist() I wasn't
removing the _objects_ from the dictionary, instead I was calling
prop_object_release() with the keysym returned by
prop_object_iterator_next()... obviously this wasn't correct.

Don't forget to free sme_evdrv in non error case too. Noticed and
reviewed by rmind.

Use only one single condvar(9) for devices and events, they are protected
by the same mutex (sme_mtx) and there's no point in using two of them.

Also add a comment mentioning some locking notes.

Reviewed and ok by rmind.

Disable debug, committed accidentally in previous.

sysmon_envsys_register: only call sme_event_unregister_all when any
other error than EEXIST is returned.

- Use only one global mutex for all tasks with devices & events.
- sysmon_envsys_register: use a SLIST to maintain allocated sme_event_drv_t
  structs for later use in sysmon_task_queue_sched(). This avoids a
  locking error acquiring/dropping the mutex multiple times.

Suggested by rmind.

Remove another COMPAT_40 block, found by rmind@.

- Build unconditionally compatibility code, there's no point to have it
  disabled, because some drivers depend in the old code yet.
- Use a new mutex for sysmon_envsys_next_sensor_index (used in
  compatibility code), otherwise bad things happen with LKMs. Thanks
  to this, the hack for LKMs has been removed.
- Check in advance if the driver doesn't exist already on the list
  before adding the sensors in the dictionary.
- Don't forget to call sme_event_unregister_all() if
  sysmon_envsys_unregister() fails after adding the array into the global
  dictionary or when creating sensors.
- Modify and add some DPRINTFs.

The lm(4) lkm works without known problems registering and unregistering
it multiple times.

- Remove ENVSYS_GSTRING and the genstr member from envsys_data_t.
  (at least three or four persons were against it).
- Add a new sensor type: ENVSYS_BATTERY_STATE, this uses value_cur
  and some predefined values in a static table, like ENVSYS_DRIVE.
- Move all static tables to sysmon_envsys_tables.c and use a function
  on it to retrieve a pointer to the struct of the specified type.
- Rename the ENVSYS_FMONDRVSTCHANGED to ENVSYS_FMONSTCHANGED and make
  it generic for Battery state and drive sensors (this flag enables
  monitoring on these sensors when state has been changed).
- Update sysmon_penvsys_event() to report state changes on
  ENVSYS_BATTERY_STATE sensors and remove other type of events, with
  PENVSYS_EVENT_STATE_CHANGED they are not necessary anymore.

Sync with HEAD.

sync with head.

Sync with HEAD.

Add a new sensor type to sysmon_envsys(9): ENVSYS_GSTRING.

ENVSYS_GSTRING (aka Generic String) uses the genstr member in
the envsys_data_t struct to add a generic string that envstat(8)
will show as value or state.

It's like the ENVSYS_DRIVER, but doesn't use value_cur. Below
is the dictionary created on these sensors:

<dict>
	<key>description</key>
	<string>acpibat0 charge state</string>
	<key>generic-state-string</key>
	<string>NORMAL</string>
	<key>monitoring-supported</key>
	<false/>
	<key>state</key>
	<string>valid</string>
	<key>type</key>
	<string>Generic string</string>
</dict>

Note that it's limited to 32 chars, but we can grow it if needed.

envstat(8) will print ENVSYS_GSTRING sensors as:

$ envstat -dacpibat0 -s"acpibat0 charge state"
  acpibat0 charge state:     NORMAL
$

- Fix a memleak in ENVSYS_SETDICTIONARY if there wasn't any error
  (thanks god for KMEMSTATS).
- sysmon_envsys_register: add all objects in the dictionary without any
  lock, at this point the sme device hasn't been added into the list
  and it's safe.
- Add sysmon_envsys_destroy_plist(prop_array_t) that destroys all objects
  associated with a device and use it on sysmon_envsys_unregister() and
  sysmon_envsys_register() if there's any error.

Thanks to Mindaugas Rasiukevicius (rmind@) for the great comments/ideas.

- sme_unregister_all: there's no need to use a struct sysmon_envsys as
  argument, just pass sme->sme_name to it.
- sysmon_envsys_register: drop the array in all cases, not just when
  there's an error. The reference is stored in the dictionary anyway...
- Update some comments.

Merge sme_event_add() into sme_event_register()... there's no need
to have two different functions for this.

Add a few KASSERT(mutex_owned(&foo_lock)) as suggested by Andrew Doran.

More misc changes for sysmon_envsys(9):

- Add the SLIST for sensor descriptions and sme_uniqsensors into the
  struct sysmon_envsys (it's per device now).
- Use only one common struct with three members for the static tables
  (there's no need to have different structs just for them).
- While initializing/destroying the events framework, use the
  strategy specified by Andrew Doran in:
  http://mail-index.netbsd.org/tech-kern/2007/06/21/0025.html).

Simplify sysmon_envsys_register() and fix an obvious error in
previous.

Some changes to improve locking on sysmon_envsys(9):

- Remove sme_mtx, a global lock (sme_list_mtx) is used to access to
  the sysmon envsys device.
- Allocate memory with KM_NOSLEEP rather than KM_SLEEP if there's a
  mutex held, to avoid sleeping.
- Remove sysmon_envsys_createplist() and add the logic into
  sysmon_envsys_register().
- sysmon_envsys_register: allocate the array and dictionaries required
  in advance for a device before the locking and adding the objects
  into the array happens.
- Rename sme_make_dictionary() to sme_add_sensor_dictionary() and pass
  to it the dictionary on which the objects will be stored for a sensor.
- Improve locking here and there.

Thanks to Mindaugas Rasiukevicius and Andrew Doran for comments.

Sync with HEAD.

Sync with HEAD.

Sync with HEAD.

In ENVSYS_SETDICTIONARY if sysmon_envsys_find() returns NULL, do not
call sysmon_envsys_release with a NULL argument.

file sysmon_envsys.c was added on branch matt-mips64 on 2007-08-05 23:16:26 +0000

In ENVSYS_SETDICTIONARY if sysmon_envsys_find() returns NULL, do not
call sysmon_envsys_release with a NULL argument.

- Do not call sme_gtredata in sysmon_envsys_createplist, as in some
  systems the interrupts are not configured and causes unexpected behaviour
  in acpi_bat and other drivers, reported by cube@.
- Update the 'type' object in sme_update_dictionary(), so that it can
  report the correct units type in acpibat(4).

Change locking stuff in sysmon_envsys_register() and
sysmon_envsys_createplist().

- ENVSYS_SETDICTIONARY: use sysmon_envsys_release() if there's an error.
- sme_events_worker: use sme_list_mtx when accessing to the sysmon_envsys
  linked list.
- Improve the comments in sysmon_envsysvar.h about the mutexes and condvar.

- sysmon_envsys_createplist: do not return the error of sme->sme_gtredata
  in all cases to the caller, just continue and show a debug printf if it
  fails, because it's not a critical error.

- Improve some debugging printfs.

Fixes another part of yamt@'s PR kern/36673, that I forgot to commit
yesterday.

Fix the last point reported by yamt in PR kern/36673:

- There must be three mandatory objects in a sensor to be valid and be
  enabled on its dictionary: units, state and desc. If any of these
  objects fails for some reason, its dictionary won't created in the
  array and sensor will be marked with ENVSYS_FNOTVALID in edata->flags.

  That means there won't never be a partial dictionary.

- Mark all invalid sensors and sensors with duplicate description with
  ENVSYS_FNOTVALID and remove ENVSYS_FDUPDESC. The former is the generic
  way to say that a sensor does not obey the rules and must be not
  created.

- sysmon_envsys_createplist: if the array does not contain any object,
  release the array and return the error.

And with all that stuff I can go to holidays and be happy.

- Do not touch sme_nsensors when sensors with duplicate description are
  found, rather maintain a global counter for all sensors that are unique,
  and use this global to get the correct index in the array.
- Use prop_array_set() rather than prop_array_add() for adding dictionaries
  into our global array. Thanks to yamt@.
- Overwrite the sensor index in the first sensor of a device, to avoid
  a problem with LKMs where the first sensor gets the index of all edata
  structures allocated in kernel.
- Add missing proto for sme_register_sensorname().

All them fix another part of yamt@'s PR.

Use the mutex that belongs to the specific task and do not mix them
blindly.

	- sme_list_mtx to access to the linked lists sme_list/sme_names_list.
	- sme_mtx to access to the data in the sysmon envsys device.

sme_update_dictionary: if there's a sensor with duplicate description,
there's no need to decrease 'i' again, just sme->sme_nsensors is enough.

Reported by yamt@.

Simplify previous.

Skip sensor with duplicate descriptions in the COMPAT_40 case, verified
with envstat(8) from netbsd-4 and sys/lkm/misc/envsys2.

Fixes another part of kern/36673 by yamt@.

- Add sme_events_destroy() to stop/destroy the callout and workqueue if
  there are no more events.
- Add sme_event_unregister_all() to remove all events associated with
  a device and use it in sysmon_envsys_unregister() to fix a "use-after-free".
- Check return error of sme_sensor_upfoo() in sme_make_dictionary() and
  return it rather than 0 all the time.
- Add more KASSERTS.

Some fixes for kern/36673 by yamt@:

- sme_register_sensorname: there's no need to handle empty list case.
- 'obj' argument in sme_sensor_upstring and similar functions is useless:
  removed the obj argument from the functions, it's handled in the function
  itself now.
- check for more errors in sme_make_dictionary() and
  sysmon_envsys_createplist().

Re-add a strlcpy() to copy the new description in edata->desc, if the
object in dictionary was updated succesfully.

Add another missing break if updating the drive-state object fails.

Add a missing break if error when setting a new description.

Replace the macros to create/update objects in dictionary with three
inline functions, suggested by yamt@.

This also gets rid of a few gotos.

sysmon_envsys_unregister: do not kmem_free(9) with the mutex held.

sme_register_sensorname: don't forget to kmem_free(9) before returning
EEXIST.

Maintain a SLIST per device to check for unique descriptions and do not
create any dictionary if it's duplicated. Only the first sensor that has
a unique description will be added.

Remove all elements from the list at sysmon_envsys_unregister() time and
free resources associated with it.

- More changes related to locking.
- Add a condvar(9) to protect deletions on the list if there's an event
  in the worker thread.

sme_make_dictionary: if prop_array_add() fails, don't forget to release
the dictionary. Pointed out by yamt@.

Add more debugging printfs to catch sensors with duplicate description,
and some cosmetic ones.

Skip sensors with the same description on the same device at attach time,
if the description of it is not unique, mark the sensor with
ENVSYS_FDUPDESC, return EEXIST and continue.

Also, when changing a description, check that the new description is not
assigned by any other sensor, otherwise return EEXIST.

Note that these restrictions are only per device. Fixes PR kern/36660.

Only allow to use ENVSYS_SETDICTIONARY if the user has write permission
in the /dev/sysmon device.

Use O_RDONLY for ENVSYS_GETDICTIONARY and O_RDWR for ENVSYS_SETDICTIONARY
in envstat(8).

Fixes PR kern/36661 by YAMAMOTO Takashi.

There's no reason why integer sensors were disabled for monitoring,
so enable it. This will allow to set a min/max limit for the
macppc/battery.c driver, which uses an integer sensor to report the
battery capacity.

Indicator/Integer sensors will share the /etc/power/scripts/sensor_indicator
script to report the events.

Sync with head.

Sync with head.

Sync with head.

Fix the old ENVSYS_GTREDATA ioctl if there are sensors with invalid
state. Enable all sensors by default as valid and current valid data and
if there is one that is in invalid state, mark current data as not valid.

This matches the behaviour with the old API. Also, this fixes the problem
with ipmi(4) reporting bogus values with sensors in invalid state using
the old envstat(8).

Note that this only happened when using the old envstat(8).

Add PENVSYS_TYPE_INDICATOR to handle events on Indicator sensors.

ipmi(4) uses this for the Intrussion and Power Supply sensors.

Imported envsys 2, a brief description of the new features:
(Part 1: API)

* Support for detachable sensors.
* Cleaned up the API for simplicity and efficiency.
* Ability to send capacity/critical/warning events to powerd(8).
* Adapted all the code to the new locking order.
* Compatibility with the old envsys API: the ENVSYS_GTREINFO
  and ENVSYS_GTREDATA ioctl(2)s are supported.
* Added support for a 'dictionary based communication channel' between
  sysmon_power(9) and powerd(8), that means there is no 32 bytes event
  size restriction anymore.
* Binary compatibility with old envstat(8) and powerd(8) via COMPAT_40.
* All drivers with the n^2 gtredata bug were fixed, PR kern/36226.

Tested by:

blymn: smsc(4).
bouyer: ipmi(4), mfi(4).
kefren: ug(4).
njoly: viaenv(4), adt7463.c.
riz: owtemp(4).
xtraeme: acpiacad(4), acpibat(4), acpitz(4), aiboost(4), it(4), lm(4).

Sync with HEAD.

Kill caddr_t; there will be some MI fallout, but it will be fixed shortly.

Back out previous for now. There isn't currently an easy way to set up
shared state before these get attached or used.

From xtraeme@.

Replace SYSMON_ENVSYS_{LOCK,UNLOCK} with mutex_{enter,exit}.

* Replace simple_locks and exclusive locks with adaptive mutexes.
* Replace ltsleep/wakeup pairs with condition variables
  (cv_init/cv_wait/cv_broadcast).

Reviewed by ad@.

sync with head.

sync with head.

Sync with head.

__unused removal on arguments; approved by core.

sync with head

- sprinkle __unused on function decls.
- fix a couple of unused bugs
- no more -Wno-unused for i386

sync with head

sync with head

Merge from HEAD.

expose the ioctl internal function now used by the acpi_apm code.

sync with head.

Sync with head.

sync with head.

sync with head.

Merge 2006-03-28 NetBSD-current into the "peter-altq" branch.

Don't attempt to call (*sme->sme_gtredata)() or (*sme->sme_streinfo)()
if those members are NULL.
As discussed on tech-kern in October/November 2005.

merge ktrace-lwp.

Fix the sync with head I botched.

Sync with HEAD.

Sync with HEAD

do wakeup() only when someone is waiting for us.

introduce SME_FLAG_BUSY and set it during operations
instead of keeping sysmon_envsys_list_slock spinlock held
because some drivers might sleep in sysmon_envsys ops,

XXX sysmon_envsys_lock is now redundant

Apply the aborted ktrace-lwp changes to a specific branch.  This is just for
others to review, I'm concerned that patch fuziness may have resulted in some
errant code being generated but I'll look at that later by comparing the diff
from the base to the branch with the file I attempt to apply to it.  This will,
at the very least, put the changes in a better context for others to review
them and attempt to tinker with removing passing of 'struct lwp' through
the kernel.

Back out the lwp/ktrace changes. They contained a lot of colateral damage,
and need to be examined and discussed more.

Pass lwp pointers throughtout the kernel, as required, so that the lwpid can
be inserted into ktrace records.  The general change has been to replace
"struct proc *" with "struct lwp *" in various function prototypes, pass
the lwp through and use l_proc to get the process pointer when needed.

Bump the kernel rev up to 1.6V

Sync with HEAD.

add envsys support to acpi A/C adapter and batteries.  This is still a work
in progress, much like the rest of ACPI.

Catch up to -current.

More catchup.

Sync kqueue branch with -current.

The device close routine is called when _last_ descriptor to the
device file is closed, rather on every descriptor close. Instead of
getting the exclusive lock on open and releasing on close, get
the lock only when needed in sysmonioctl_envsys().
Fixes kern/14368 by Anthony Mallet. Okayed by Jason Thorpe.

While here, also inline sysmon_envsys_init() to sysmonopen_envsys(),
and g/c sysmonioctl_wdog() prototype.

Catch up to -current.

add RCSIDs

Various changes to make cloning devices possible:

	* Add an extra argument (struct vnode **) to VOP_OPEN. If it is
	  not NULL, specfs will create a cloned (aliased) vnode during
	  the call, and return it there. The caller should release and
	  unlock the original vnode if a new vnode was returned. The
	  new vnode is returned locked.

	* Add a flag field to the cdevsw and bdevsw structures.
	  DF_CLONING indicates that it wants a new vnode for each
	  open (XXX is there a better way? devprop?)

	* If a device is cloning, always call the close entry
	  point for a VOP_CLOSE.


Also, rewrite cons.c to do the right thing with vnodes. Use VOPs
rather then direct device entry calls. Suggested by mycroft@

Light to moderate testing done an i386 system (arch doesn't matter
though, these are MI changes).

Sync with HEAD.

file sysmon_envsys.c was added on branch thorpej_scsipi on 2000-11-22 16:04:54 +0000

Split the envsys and watchdog parts of sysmon into their own
files.